A key to univariate statistics (version 0.3)

This key to univariate statistics is by no means complete but just a selection of frequently used models. The use of each model is explained through examples including syntax from the R programming language. The R-syntax imports data directly from the web, so you don't have to import your own data to follow the examples.

If you are unfamiliar with R, you may have a look at our Mini-Intro, or go to the home page for the R-project, http://www.r-project.org, where you find all documentation you need to get started.

Look for properties of your response variable to find a suitable test:

Response variable is continuous Response variable represents counts or frequencies
The test to chose depends on properties of your predictor variable(s) as well:
  • No predictor variables, go to One sample tests.
  • Continuous predictor variable(s), go to Regression.
  • Only categorical predictor variable(s), go to Two sample tests, or ANOVA.
  • Both continuous and categorical predictor variables, go to ANCOVA.
  • You have clustered data where the factor representing clusters is considered to be a random effects variable (e.g. repeated measurements, block designs, split-plot designs, factorial designs, longitudinal data, line transects, etc.) go to Linear mixed-effects models.
  • If you have a response variable with count data, or the response variable represents frequencies of different categories, go to Log linear models and frequency tables.
    Response variable represents proportions or is binary Response variable represents time at occurrence
    If you have a response variable representing proportions or binary data, go to Logistic regression. If you have time-at-death data, or data that measures the time it takes for individual observations to reach a certain status (e.g. germination of seeds), go to Survival analysis.

    This key was last updated on: Tuesday 2 September 2008 by Knut Helge Jensen